AI Can Help Indigenous People Protect Biodiversity

The United Nation’s latest biodiversity report warns that one million species are headed for extinction, including 1,000 wild mammal species and 450 species of sharks and rays. Losing this much wildlife will have massive human consequences because one in five people depend on these species for food and income. To help reverse the crisis, the UN’s Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services urges governments and NGOs to secure land rights for Indigenous peoples, who have been “proactively” protecting wild species with “local and regional monitoring networks.”

Given my research on the symbiosis between human and nonhuman intelligence, I’m curious to see what role technology might play in serving these twin goals: conserving biodiversity and securing Indigenous land rights. In particular, an advanced form of artificial intelligence called deep-learning neural networks, able to improve its performance without being programmed, is revolutionizing the analysis of sound data from at-risk ecosystems. This acoustic monitoring, often complementing visual monitoring and sometimes replacing it, is conservation for the 21st century. Restricted access, of course, remains a perennial problem with AI—both in code-sharing and in the use of the technology itself. Governments and NGOs need to work harder to make this tool available to rapid-response teams, otherwise its promise would be wasted.

Deep-learning neural networks don’t negate the need for human vigilance; in fact, the opposite is true. Because this advanced AI can analyze sound data within seconds, it opens up a window of time for critical intervention, especially in places plagued by illegal logging and mining. The technology could hear the sounds of chainsaws or drills, and send alerts to action-ready patrols on the ground. Indigenous communities with access to this technology and able to respond instantly could be a key part of this new equation.

One of the first AI-based Indigenous conservation projects, undertaken by Cornell University, was co-developed with the Coral Gardeners, from Mo’orea, French Polynesia. Founded in 2017, this Indigenous group cultivates heat-resistant super corals and transplants them onto damaged parts of the reef. Cornell provides the software to track the sounds of the many organisms making their home here and, working also with the University of Hawaii, integrates them into a recording platform, ReefOS, a network of sensors and cameras collecting visual and acoustic data 24 hours a day. The AI-mediated soundscape tells the on-site respondents whether the reefs are starting to sound like healthy and stable reef systems, or whether additional restoration efforts are needed.

Such Indigenous-AI conservation works just as well with other ecosystems. In March 2014, the Tembé tribe in Northern Brazil reached out to the San Francisco nonprofit Rainforest Connection to build a low-cost alert system to monitor deforestation. Rainforest Connection uses recycled cell phones and an open-source AI software called TensorFlow to single out the sounds of chainsaws and logging trucks amid the cacophony of the Amazon. Text alerts go out instantly to Tembé patrols when Google’s cloud computing detects the rev of a chainsaw.

Unfortunately, this cutting-edge partnership is thrown into jeopardy by rampant encroachment on Indigenous lands, led by local politicians eager to clear the forests for industrial farming and ranching. One such politician is João Cleber Torres, mayor of São Félix do Xingu, a city that routinely posts some of the highest deforestation rates in Brazil, and where unsolved, land-related murders are just as routine.

With only about 2,000 members, the Tembé have found it difficult to fend off hired guns laying siege to their 2,766 square kilometers of ancestral holdings. Already 30 percent of their forests have been destroyed. There are human casualties as well. Under President Jair Bolsonaro, such Indigenous killings have skyrocketed in Brazil. On April 14, 2021, the Vatican published an emergency declaration from the Coordinator of Indigenous Organizations of the Amazon River Basin, warning that 202 Indigenous leaders had been murdered in 2020—averaging more than one death every other day—done with impunity by land-grabbers protected by their political bosses.

The consequences for biodiversity are dire. Throughout the world, not just in the Amazon but also in other species-rich ecosystems like the Congo Basin, assault on Indigenous lands has decimated once-flourishing habitats, exposing them to extractive industries at their most lawless. Without legally secured land rights, the partnerships necessary for AI-mediated conservation are impossible.

That irony is likely to persist under the World Wildlife Fund’s 30x30 plan—to conserve 30 percent of Earth by 2030. This is being negotiated for the final draft of the U.N. Convention on Biological Diversity, to take place in Montreal this December. The top-down target could end up evicting as many as 300 million human forest-dwellers. Biden’s separate 30x30 plan, though hardly perfect, at least offers better aspirational goals. The U.S. government will respect “Tribal sovereignty.” It will “leverage resources with Tribes,” “incorporate Indigenous Traditional Ecological Knowledge into decision-making,” and “take an inclusive and collaborative approach to the stewardship of land and water resources.”

Artificial intelligence is an integral part of this partnership. Especially telling was the presence of the Allen Institute for AI at this year’s Our Ocean Conference, co-hosted by the U.S. and the Republic of Palau, bringing Indigenous-AI conservation to a whole new level. The conference showed that AI can perform specific tasks beneficial to island nations, like tracking illegal fishing that depletes fish stocks and threatens local livelihood. More generally, AI, if carefully designed and rigorously tested before being deployed in the field, can be a useful tool for analyzing ocean data, from the songs of humpback whales to the properties of microplastics. The $20 million commitment from the National Science Foundation to create an AI Institute for climate and coastal oceanography raises a lot of hopes. However, competing demands for these funds also suggests that it won’t all be smooth sailing. It’s going to take some doing to integrate Indigenous conservation with the ocean-based blue economy that’s clearly driving the research agenda.

The Palau Conference set a very high bar. Unfortunately, the Biden administration isn’t always able to stick to its principles. Two months after Our Ocean 2022, at the Summit of the Americas—which avoided any mention of the devastated rainforest or murders in the Amazon to secure Bolsonaro’s presence—Indigenous leaders critical of him were barred from entry. An incident like this shows that the rhetoric of partnership is often just that: rhetoric. Biden’s commitment to biodiversity also seems shaky under the otherwise monumental Inflation Reduction Act, given the mandated, decade-long leasing of 60 million acres of off-shore waters and 2 million acres of public lands for oil and gas drilling. A provision of $272.5 million for Native-led climate resilience and adaptation is a silver lining here. AI-mediated conservation, centered on Native communities and making their participation meaningful, is one way to keep biodiversity live, innovative, and politically consequential.

This is an opinion and analysis article, and the views expressed by the author or authors are not necessarily those of Scientific American.